Genetic algorithm for affine point pattern matching
|
|
- Clifton Archibald Weaver
- 8 years ago
- Views:
Transcription
1 Pattern Recognition Letters 24 (2003) Genetic algorithm for affine point pattern matching Lihua Zhang, Wenli Xu *, Cheng Chang Department of Automation, Tsinghua University, Beijing , PR China Received 14 September 2001; received in revised form 25 April 2002 Abstract Point pattern matching (PPM) is an important topic in the fields of computer vision and pattern recognition. According to if there exists a one to one mapping between the two point sets to be matched, PPM can be divided into the case of complete matching and the case of incomplete matching. According to if utilizing information other than 2-D image coordinates, PPM can be divided into labelled point-matching case and unlabelled point-matching case. Using partial Hausdorff distance, this paper presents a genetic algorithm (GA) based method to solve the incomplete unlabelled matching problem under general affine transformation. Since it successfully reduces the solution space of GA by constructing feature ellipses of point sets, the method can achieve high computing efficiency and good matching results. Theoretical analysis and simulation results show that the new algorithm is very effective. Ó 2002 Elsevier Science B.V. All rights reserved. Keywords: Point pattern matching; Affine transformation; Genetic algorithm; Partial Hausdorff distance; Feature ellipses 1. Introduction Point pattern matching (PPM) is an important problem in the fields of computer vision and pattern recognition. Its main task is to pair up the points in two images of a same scene. It can be widely used in many applications, such as image registration, object recognition, object tracking, autonomous navigation, pose estimation, etc. According to if there exists a one to one mapping between the two point sets to be matched, PPM can be divided into the case of complete matching and the case of incomplete matching. Up * Corresponding author. address: zhanglihua@ieee.org (L. Zhang). to now, the case of complete matching has been extensively treated and many effective algorithms (Atkinston, 1987; Griffin and Alexopoulos, 1989; Hong and Tan, 1988; Lavine et al., 1983; Scott and Longuet-Higgins, 1991; Shapiro and Brady, 1992; Simon et al., 1972; Sprinzak and Werman, 1994; Wang, 1983; Zahn, 1974; Zhang Lihua and Xu Wenli, 1999a; Zhang Lihua and Xu Wenli, 1999b) have been put forward. However, in practical applications, it is inevitable that there exist spurious or lost points in the images. So research on the case of incomplete matching is much more important. According to if utilizing additional information (such as color, intensity, etc.) other than 2-D image coordinates, PPM can be divided into labelled point-matching case and unlabelled point matching. Clearly the unlabelled point matching is more /03/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved. PII: S (02)
2 10 L. Zhang et al. / Pattern Recognition Letters 24 (2003) 9 19 difficult than the former, but it can be used in more general applications. This paper will focus on incomplete unlabelled point pattern-matching problem. Vindo and Ghose (1993) treat the pointmatching problem under Euclidean transformation as a 0 1 IPP problem and solve it by an asymmetric neural network. Morgera and Cheong (1995) solve the point-matching problem under rigid transformation by using the theory of Lie group. However, their algorithm can only be used for the case in which one point pattern is a subset of the other. Stockman et al. (1992) use a clustering approach to estimate the similarity transformation parameters and then determine the corresponding point pairs. Goshtasby and Stockman (1985) use the convex hulls of the point sets to reduce the complexity of the matching problems. Unfortunately, if there exist spurious or lost points on the convex hulls, the algorithm fails. The cluster-based approach proposed by Chang et al. (1997) is very effective. But it cannot insure finding out the optimal solutions with as many matching point pairs as possible, since it solves the problem from bottom to top. Recently, Zhang Lihua and Xu Wenli (in press) introduce some definitions of matching clique, support point pair, index set, and index matrix, etc. Based on these definitions, several properties and a theorem are proposed, and a new algorithm to solve the problem of matching two point sets with the different cardinality under rigid transformation is presented. On the contrary to other methods in the literature, the algorithm searches for the optimal solutions from top to bottom. Therefore, it can find as many matching point pairs as possible. However, the algorithm cannot be directly used under similarity transformation and affine transformation. Compared with the incomplete matching problems under Euclid (rigid) and similarity transformations, the incomplete matching problem under general affine transformation is much more difficult. Therefore, in current literatures, only a small ratio really discusses the incomplete unlabelled point-matching problem. According to the observation that any triple of points can define a valid plane affine transformation, Lamdan et al. (1988) use a voting scheme geometric hashing, which considers all possible triples of points. Obviously its computational complexity is very high. In contrast, the method proposed by Huttenlocher (1991) is able to eliminate most matches by considering just pairs of points. The remaining matches are then enumerated. The method is based on a new affine invariant constructed by distance ratios defined by quadruples of points. However, the method will fail if three pre-selected points used for defining the new affine invariant do not appear in both point sets. The current algorithms found in literature try to solve the difficult problem using very conventional techniques, e.g. method of exhaustion and geometric hashing. Undoubtedly, these conventional techniques are ever very useful for solving Euclid and similarity cases. However, for complex affine transformation case, their efficiency is not very high. Oppositely, using partial Hausdorff distance, this paper presents a method based on an unconventional technique genetic algorithm (GA) to solve the incomplete point-matching problem. The method can achieve high computing efficiency and good matching results, because it successfully reduces the solution space of GA by constructing feature ellipses of point sets. Theoretical analysis and simulation results show that the new algorithm is very effective. 2. Principle of algorithm In order to grasp our algorithm as a whole, we firstly introduce its fundamental idea briefly. Given two point sets P and P 0 which satisfy a general affine transformation relation, before constructing chromosomes, a triple of points should be chosen on the feature ellipse of P. Here the triple of points may not be the real points of P to be matched and it is named the referenced triplet. The concept of feature ellipse will be described in the section of preliminary knowledge. Each chromosome is composed of a triple of points of P 0 using binary coding. Similarly chromosomes are not requested to be real points of P 0, but points locating in the image plane of P 0. Obviously the triple of points represented by each chromosome and the referenced triplet can
3 L. Zhang et al. / Pattern Recognition Letters 24 (2003) determine a valid plane affine transformation T. Under this transformation, point pattern P is transformed to T ðpþ. Clearly, we can judge if T is near the original transformation by comparing point patterns T ðpþ and P 0. Considering the partial bidirectional Hausdorff distance (Huttenlocher et al., 1993) can measure the degree of match between two point sets without explicitly pairing points of T ðpþ and P 0, we construct fitness function by the partial bidirectional Hausdorff distance. The output of our GA is the triple of points which has the highest value of fitness function. Using the triplet and the referenced triplet, the transformation parameters can be solved to realize final point matching. Intuitively the idea of our algorithm is similar to method of exhaustion and geometric hashing, but its computational complexity is far too lower than the two methods. This is determined by two main reasons. One reason is that GA features implicit parallelism. The other one is that the initial population of our genetic is not generalized completely randomly, but generalized in the local area demarcated by the feature ellipse of point set P 0, which is equivalent to restricting the possible solution space to be searched by genetic algorithm. reasonable triplet. To find a reasonable triplet, we give the following analysis. Under the case of no outliers (including spurious points and lost points), suppose the original affine transformation, which transforms P to P 0, is T. Then the convex polygon region S 1 formed by the convex hull of P is transformed to region T ðs 1 Þ. Clearly T ðs 1 Þ should have the same position and shape with the convex polygon region S 2 formed by the convex hull of point set P 0. Therefore, points locating in S 1 will be transformed to points locating in S 1. Based on this observation, the referenced triplet can be chosen from the interior of S 1, because its corresponding matching triplet must lie in S 2, certainly in the bounded image plane of P 0. However, if outliers exist, the conclusion may be not true, because the outliers may disturb the convex hulls of the two point sets. The worst situation is the region S 2 does not contain some regions of T ðs 1 Þ. Obviously if the referenced triplet chosen in S 1 locates in these regions, it must be unreasonable. To solve this, we construct so called feature ellipse of a point set. Definition 1. (Feature ellipse of a point set) Given a point set P ¼fp i ¼ðx i ; y i ; 1Þ T ji ¼ 1; 2;...; mg, its feature ellipse is defined as 3. Feature ellipse of a point set ðx cþ T E 1 ðx cþ ¼ 1 S ; ð1þ According to the principle of our algorithm, a referenced triplet should be chosen in the image plane of P and once it is paired with a triple of points represented by each chromosome, parameters of the corresponding transformation can be determined. Thus, the physical explanation of the GA is to find the most matchable triple of points in the image plane of P 0 to the referenced triplet. Obviously, if we randomly choose any three points in the image plane of P as the referenced triplet, it is possible that no matchable triplet can be found in the bounded image plane of P 0. This is because the numerical range of the triplet, which the referenced triplet is transformed to by the original affine transformation, is beyond the scope that chromosomes can express. Such a referenced triplet is called an unreasonable triplet, or else a where S is a positive integer whose value determines the size of the feature ellipse, and c ¼ 1= m P m i¼1 p i and E ¼ 1=m P m i¼1 ðp i cþðp i cþ T are the center point and the second-order center moment of P, respectively. Clearly the center of the feature ellipse locates in the canter of the point set, and the shape of the feature ellipse approximates the shape of P, i.e. the distribution of points of P. Via adjusting the parameter S, the feature ellipse can be insured to lie in the convex hull of P (certainly in the bounded image plane of P), further in a local region around the center of P. Draw three rays from the center c of P, of which each pair forms an angle of 120. The rays intersect the feature ellipse at three points, which is the expected referenced triplet. Clearly the referenced
4 12 L. Zhang et al. / Pattern Recognition Letters 24 (2003) 9 19 triplet locates in the local area around the center of P. Here, the request that angles formed by each pair of rays be 120 is to make the referenced triplet nonlinear and interdistance between any two points large enough. It is easily proved that under the case of no outliers and no image noises, for the same S the feature ellipse of P must correspond to the one of P 0. Thus, the matching triplet of the referenced triplet can be found on the feature ellipse of P 0. This shows that the chosen referenced triplet is reasonable because the feature ellipse of P 0 must lie in the bounded image plane of P 0. Obviously the usage of the concept of feature ellipse is equivalent to restricting the search space of the GA, that is the algorithm can search for the matching triplet of the referenced triplet just on (notice here not in) the feature ellipse of P 0. Considering that there must exist image noises and outliers in practical applications, the matching triplet corresponding to the referenced triplet may not locate on the feature ellipse of P 0. However, it will lie in a local region around the ellipse. Thus, the search space of the GA is reduced to this local area. 4. GA-based algorithm Professor J. Holland put forward the essential idea of GA in GA is a highly efficient stochastic global optimization algorithm for problem solving. Beginning from any randomized initial population that consists of a group of chromosomes (i.e. potential search nodes), new populations are arrived at by using various genetic operators with subsequent iterations. A new chromosome replaces a previous one if it is judged to be better than it according to a function called fitness. Two main advantages of GA are it is not easy to slump into a local optimized region and it can be realized by a parallel computational method. GA has been applied to computer vision in many fields. Some of the works related to medical image registration, image segmentation, modelbased matching, and affine invariant recognition can be found in (Bhanu et al., 1991; Grefenstette and Fitzpatrick, 1985; Hill and Taylor, 1992; Tsang, 1997a,b). To use GA in finding point correspondences, three problems are to be solved, constructing fitness function, coding chromosomes, and choosing suitable genetic operators Coding chromosomes Given two point sets P and P 0 with m and n points respectively, each chromosome consists of a triple of points in the image plane of P 0. Its detailed construction method is seriating coordinates of the three points and coding them into binary codes. Here using binary coding is because the image resolution is always limited, for example, or , and then the coordinate of points can be represented by integers in the scope of the image resolution. Even if the location precision of points is less than a pixel, we can still obtain enough representation precision by changing the length of binary code. Given three points pj 0 0, pj 0 1 and pj 0 2 in the image plane of P 0 (may be not the points of P 0 ), Table 1 shows the detailed construction method of a chromosome consisting of the three points. Here, each x or y is represented by a binary code. Thus the whole chromosome is a binary sequence. Notice, in order for the genetic operators operating conveniently and obtaining a more precious matching pair of triplets, the chromosome is not divided into genes. The vertical lines are simply used for description. Therefore the operations of genetic operators are all executed with the unit of bit Generating initial population The initial population is generated randomly. However, the generating range of chromosomes is not arbitrary but limited to the local area around the feature ellipse of P 0. Table 1 Chromosome coding x 0 j 0 yj 0 0 x 0 j 1 yj 0 1 x 0 j 2 yj 0 2
5 L. Zhang et al. / Pattern Recognition Letters 24 (2003) Constructing fitness function Given a chromosome shown in Table 1, the coordinates of the triplet that the chromosome represents are firstly retrieved according to their binary codes. Then the parameters of the affine transformation, denoted by T, can be easily solved from this triplet and the referenced triplet. Obviously, if the degree of match between P and P 0 under the transformation can be measured, it is equivalent to evaluating the fitness of the chromosome. Considering the partial bidirectional Hausdorff distance between point sets P 0 and T ðpþ to which P is transformed by T, smaller the distance is, the degree of match between P and P 0 is larger. So the fitness function can be selected as the inverse of the partial bidirectional Hausdorff distance, fitness ¼ 1 1 þ H LK ðt ðpþ; P 0 Þ : ð2þ Notice here, the denominator is the partial bidirectional Hausdorff distance plus 1 in order to avoid zero appearing in the denominator Defining genetic operators Selection: Selection operator is used to select good chromosomes that contribute their geneinherited knowledge for the next generation. Here we use commonly used Roulette-wheel selection process. In Roulette-wheel selection process, the selection probability of each individual is proportional to its fitness value. Crossover: Although selection operator can keeps fitter individuals (i.e. chromosomes) in evolutionary process, it does not create any new individual. Crossover operator can produce new chromosomes through combining partial structure of two father individuals. Here we adopt commonly used single point crossover operator. Mutation: In order to prevent the loss of diversity in the evolutionary process, an operator named uniform mutation is designed and carried on in a small probability. It operates on each binary bit of a chromosome in another predefined probability. We realize the mutation operator by reverse the value of the current binary bit, i.e. 0 1, Full genetic algorithm Algorithm 1. Genetic algorithm for affine point matching Given two point patterns to be matched with each other P and P 0, choose population size N, crossover probability p c, mutation probability p m, two fractions f L and f K of the partial bidirectional Hausdorff distance and the maximum iterative steps G max. Step 1: Compute the feature ellipses of P and P 0, and then solve the referenced triplet in the image plane of P and determine the local search area in the image plane of P 0 for the GA. Step 2: Randomly generate N triplets in the local search area and then convert them into chromosomes for initial generation. Step 3: Compute fitness function values of all chromosomes in current population and then select fitter chromosomes by selection operator. Step 4: Apply crossover operator at the probability p c and mutation operator at the probability p m to the fitter chromosomes and generate the population of next generation. Step 5: If the maximum iterative steps G max is not reached, go to Step 3. Otherwise let T best be the affine transformation determined by the best chromosome, and match point patterns T best ðpþ and P 0 according to the simple nearest neighbor rule. 5. Experimental results To verify the effectivity of our algorithm, a lot of simulated experiments are performed. What follows is an example of the simulations. Firstly a point pattern P with 15 points is generated randomly in a numerical bound of , and another point pattern P 0 is obtained after an affine transformation with
6 14 L. Zhang et al. / Pattern Recognition Letters 24 (2003) 9 19 A ¼ 0 1:5 0:9 0 and t ¼½10 10Š T is applied to P. After the first point is deleted from P 0, two randomly generated spurious points are added to P 0. Finally Gaussian noises with mean-value zero and variance 1.0 are added to coordinates of each point in P and P 0. Here noise range is bounded to less than 3. The final point patterns P and P 0 are shown in Figs. 1 and 2, respectively. And the coordinates of each point are listed in Tables 2and 3. Fig. 1. Point pattern P. Fig. 2. Point pattern P 0. In the experiment, the parameters of Algorithm 1 are set as, population size N ¼ 400, crossover probability p c ¼ 0:05, mutation probability p m ¼ 0:02, two fractions f L and f K are all 0.7, and the maximum iterative steps G max ¼ 100. The length of binary code for each coordinate is 16 and the parameter S in Eq. (1) is 4. After randomly running Algorithm 1 for 10 times, the best chromosome is obtained and its corresponding affine transformation parameters are 0:0031 1:4874 A ¼ ; 0:8663 0:0073 t ¼ ½15: :4039 Š T : Under this transformation, point pattern P is transformed to T ðpþ. To give an intuitive observation, we plot T ðpþ together with P 0 in Fig. 3. It is easily seen that each matchable point approach its pairing point very near. Using the simple nearest neighbor rule, the matching result shown in Table 4 is obtained. Clearly the result is completely correct. To further demonstrate Algorithm 1 s adaptability, here we briefly described another experiment in which a larger point set P with 50 points is randomly generated. Point pattern P 0 is obtained after a random affine transformation with P. A ¼ 0:2967 0:2683 0:4695 0:5192 ; t ¼ ½ Š T is applied to P. Then we randomly deleted 20 points from P 0. Meanwhile, 20 randomly generated spurious points are added to P 0. Finally, larger Gaussian noises with mean-value zero and variance 5.0 are applied. Clearly, the experiment condition is much more rigorous than the previous experiment. The final point patterns P and P 0 are shown in Figs. 4 and 5, respectively, and the coordinates of each point are listed in Tables 5 and 6. During the experiment preparing process, since we could know which points are deleted and which points are added, the expected matching result can be predicted in Table 7. The experimental parameters of Algorithm 1 are set as, N ¼ 500, p c ¼ 0:03, p m ¼ 0:03, f L and f K are all 0.6, and G max ¼ 100. The length of binary
7 L. Zhang et al. / Pattern Recognition Letters 24 (2003) Table 2 Coordinates of point pattern P x y Table 3 Coordinates of point pattern P x y Fig. 3. Point pattern T best ðpþ and P 0. with the expected result listed in Table 7, clearly we correctly matched 25 of 30 matching pairs, the other two matching pairs are incorrect because the spurious points are too close to the original points. So the simple nearest neighbor rule cannot differentiate them from the correct matchings. In fact, the matching result can be further improved by utilizing more advanced optimized matching techniques (Chang et al., 1997; Xu Wenli and Zhang Lihua, 2001) for point pattern matching under rigid transformation instead of the simple nearest neighbor rule. This is because the matching subsets of point pattern T ðpþ and P 0 should satisfy a rigid transformation relation instead of an affine transformation. code for each coordinate is still 16 and the parameter S is also 4. After randomly running Algorithm 1 for 10 times, the best chromosome is obtained and its corresponding affine transformation parameters are 0:3264 0:2543 A ¼ ; 0:4634 0:5152 t ¼ ½287: :1693 Š T : Clearly this result is very close to the true affine transformation used to setup the experiment. Under this transformation, point pattern P is transformed to T ðpþ. Similarly, to give an intuitive observation, T ðpþ is plotted together with P 0 in Fig. 6. Using the simple nearest neighbor rule, a matching result is obtained in Table 8. Compared 6. Complexity analysis Supposing the size of the two point sets to be the same n, determining the complexity of the algorithm proposed in this paper is very straightforward: n 2 N G, where Oðn 2 Þ is the complexity of computing the Hausdorff distance between the two point sets, N is the size of the population and G is the number of iterative generations. Here N and G are preset parameters and they are relatively independent of n. In our experiments, N varies from 100 to 1000, and G is adjusted from 60 to 100. Now we turn to discuss the complexity of an exhausting method. If using the same measurement (Partial Hausdorff distance), for three given points (i.e. a referenced triplet) in point set P 0,
8 16 L. Zhang et al. / Pattern Recognition Letters 24 (2003) 9 19 Table 4 Result of point pattern matching P P Fig. 4. Point pattern P in experiment 2. Fig. 5. Point pattern P 0 in experiment 2. Table 5 Coordinates of point pattern P x y x y Table 6 Coordinates of point pattern P x y x y a randomized algorithm which exhausts all the possible matching triplets in point set P would have complexity of n 2 Cn 3. An implicit big assumption here is that the given referenced triplet in P 0 should
9 L. Zhang et al. / Pattern Recognition Letters 24 (2003) Table 7 Expected result of point pattern matching P P P P Fig. 6. Point pattern T best ðpþ and P 0 in experiment 2. Table 8 Testel result of point pattern matching P P P P also appear in the other point set P. Obviously this assumption might not be true in case of incomplete point matching. Suppose the probability of a point appearing in both P 0 and P is r. Easily verified, to guarantee finding a referenced triplet appearing in both point sets, the least combination number of referenced triplets that should be tried is ðc 1 nð1 rþ C2 nr þ C2 nð1 rþ C1 nr þ C3 nð1 rþ Þ: Therefore, the total computation complexity of an exhausting method would be n 2 C 3 n ðc1 nð1 rþ C2 nr þ C2 nð1 rþ C1 nr þ C3 nð1 rþ Þ; i.e. Oðn 8 Þ. Comparing Oðn 8 Þ and n 2 N G, it is easily concluded that with the increase of n, our GAbased algorithm s computation efficiency is much better than the exhausting method. For instance, in our second experiment, the parameters we used are n ¼ 50, N ¼ 500, G ¼ 100 and r ¼ 0:60. Easily know the complexity of the GA is around 1:2e þ 8, but the complexity of an exhausted algorithm will
10 18 L. Zhang et al. / Pattern Recognition Letters 24 (2003) 9 19 be nearly 1:2e þ 13. So Algorithm 1 reduced the computational complexity to 1:0e 5 of the exhausted method with n ¼ 50. This further proves the effectivity of our algorithm. 7. Conclusions In this paper, using partial Hausdorff distance, a GA-based method for the incomplete point pattern-matching problem under general affine transformation is presented. The method can achieve high computing efficiency and good matching results, because it successfully reduces the solution space of GA by constructing feature ellipses of point sets. Theoretical analysis and simulation results show that the new algorithm is very effective. Before the paper is finalized, our future research direction will be briefly mentioned. That is, although the proposed algorithm is very effective in normal case, if the referenced triplet of point set P is wrongly chosen, there will be no corresponding triplet in the local search area of P 0. To solve this problem, a simple possible solution is to expand the search area in P 0. However, this may decrease the algorithm s efficiency to some extent. Therefore, our future research will focus on issuing a better solution. Acknowledgements This work was supported by the 985 project of Tsinghua University and the Major Research Project of the tenth-five plan of PR China (no. 2001BA609A). References Atkinston, M.D., An optimal algorithm for geometrical congruence. J. Algorithms 8, Bhanu, B., Lee, S., Ming, J., Self-optimizing image segmentation system using a genetic algorithm. In: Proceedings of the Fourth International Conference on Genetic Algorithms, San Diego, CA, pp Chang, S.H., Cheng, F.H., Hsu, W.H., Wu, G.Z., Fast algorithm for point pattern matching: invariant to translations rotations and scale changes. Pattern Recogn. 30 (2), Goshtasby, A., Stockman, G.C., Point pattern matching using convex hull edges. IEEE Trans. Syst. Man Cybernet. SMC 15 (5), Grefenstette, J.J., Fitzpatrick, J.M., Genetic search with approximate function evaluation. In: Proceedings of International Conference on Genetic Algorithms and their Applications, pp Griffin, P.M., Alexopoulos, C., Point pattern matching using centroid bounding. IEEE Trans. Syst. Man Cybernet. 19 (5), Hill, A., Taylor, C.J., Model-based image interpretation using genetic algorithm. Image Vision Comput. 10, Hong, J.W., Tan, X.N., A new approach to point pattern matching. In: Proceedings of Ninth International Conference on Pattern Recognition, pp Huttenlocher, D.P., Fast affine point matching: an output-sensitive method. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, pp Huttenlocher, D.P., Klanderman, G.A., Rucklidge, W.J., Comparing images using the Hausdorff distance. IEEE Trans. Pattern Anal. Mach. Intell. 15 (9), Lamdan, Y., Schwartz, J.T., Wolfson, H.J., Object recognition by affine invariant matching. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, pp Lavine, D., Lambird, B.A., Kanal, L.N., Recognition of spatial point patterns. Pattern Recogn. 16 (3), Morgera, S.D., Cheong, P.L.C., Rigid body constrained noisy point pattern matching. IEEE Trans. Image Process. 4 (5), Scott, G.L., Longuet-Higgins, H.C., An algorithm for associating the features of two images. Phil. Trans. Roy. Soc. London B 244, Shapiro, L.S., Brady, J.M., Feature-based correspondence: an eigenvector approach. Image Vision Comput. 10 (5), Simon, J., Checroun, A., Roche, C., A method of comparing two patterns independent of possible transformations and small distortions. Pattern Recogn. 4 (1), Sprinzak, J., Werman, M., Affine point matching. Pattern Recogn. Lett. 15 (4), Stockman, G., Kopstein, S., Benett, S., Matching images to models for registration and object detection via clustering. IEEE Trans. Pattern Anal. Mach. Intell. PAMI 4 (3), Tsang, P.W.M., 1997a. A genetic algorithm for affine invariant object shape recognition. In: Proceedings of First IEE/IEEE International Symposium on Genetic Algorithm in Engineering Systems, GALESIA, pp Tsang, P.W.M., 1997b. A genetic algorithm for affine invariant recognition of object shapes from broken boundaries. Pattern Recogn. Lett. 18, Vindo, V.V., Ghose, S., Point matching using asymmetric neural networks. Pattern Recogn. 26 (8),
11 L. Zhang et al. / Pattern Recognition Letters 24 (2003) Wang, C., Some experiments in relaxation image matching using corner features. Pattern Recogn. 16 (2), Xu, Wenli, Zhang, Lihua, A geometric reasoning based algorithm for point pattern matching. Science in China, Series F 44, Zahn, C.T., An algorithm for noisy template matching. In: Proceedings of the IFIP Congress, pp Zhang Lihua, Xu Wenli, in press. New algorithms for pointpattern matching using index matrix. In: The Third Asian Control Conference. Zhang, Lihua, Xu, Wenli, 1999a. Point-pattern matching using irreducible matrix and relative invariant. Tsinghua Sci. Technol. 4 (4), Zhang, Lihua, Xu, Wenli, 1999b. Point pattern matching. China J. Comput. 22 (7),
CS 534: Computer Vision 3D Model-based recognition
CS 534: Computer Vision 3D Model-based recognition Ahmed Elgammal Dept of Computer Science CS 534 3D Model-based Vision - 1 High Level Vision Object Recognition: What it means? Two main recognition tasks:!
More informationThresholding technique with adaptive window selection for uneven lighting image
Pattern Recognition Letters 26 (2005) 801 808 wwwelseviercom/locate/patrec Thresholding technique with adaptive window selection for uneven lighting image Qingming Huang a, *, Wen Gao a, Wenjian Cai b
More informationGenetic Algorithms commonly used selection, replacement, and variation operators Fernando Lobo University of Algarve
Genetic Algorithms commonly used selection, replacement, and variation operators Fernando Lobo University of Algarve Outline Selection methods Replacement methods Variation operators Selection Methods
More informationA Robust Method for Solving Transcendental Equations
www.ijcsi.org 413 A Robust Method for Solving Transcendental Equations Md. Golam Moazzam, Amita Chakraborty and Md. Al-Amin Bhuiyan Department of Computer Science and Engineering, Jahangirnagar University,
More informationGenetic Algorithm Evolution of Cellular Automata Rules for Complex Binary Sequence Prediction
Brill Academic Publishers P.O. Box 9000, 2300 PA Leiden, The Netherlands Lecture Series on Computer and Computational Sciences Volume 1, 2005, pp. 1-6 Genetic Algorithm Evolution of Cellular Automata Rules
More informationA Service Revenue-oriented Task Scheduling Model of Cloud Computing
Journal of Information & Computational Science 10:10 (2013) 3153 3161 July 1, 2013 Available at http://www.joics.com A Service Revenue-oriented Task Scheduling Model of Cloud Computing Jianguang Deng a,b,,
More informationBuild Panoramas on Android Phones
Build Panoramas on Android Phones Tao Chu, Bowen Meng, Zixuan Wang Stanford University, Stanford CA Abstract The purpose of this work is to implement panorama stitching from a sequence of photos taken
More informationA Non-Linear Schema Theorem for Genetic Algorithms
A Non-Linear Schema Theorem for Genetic Algorithms William A Greene Computer Science Department University of New Orleans New Orleans, LA 70148 bill@csunoedu 504-280-6755 Abstract We generalize Holland
More informationA Simple Feature Extraction Technique of a Pattern By Hopfield Network
A Simple Feature Extraction Technique of a Pattern By Hopfield Network A.Nag!, S. Biswas *, D. Sarkar *, P.P. Sarkar *, B. Gupta **! Academy of Technology, Hoogly - 722 *USIC, University of Kalyani, Kalyani
More informationTracking Moving Objects In Video Sequences Yiwei Wang, Robert E. Van Dyck, and John F. Doherty Department of Electrical Engineering The Pennsylvania State University University Park, PA16802 Abstract{Object
More informationThe Role of Size Normalization on the Recognition Rate of Handwritten Numerals
The Role of Size Normalization on the Recognition Rate of Handwritten Numerals Chun Lei He, Ping Zhang, Jianxiong Dong, Ching Y. Suen, Tien D. Bui Centre for Pattern Recognition and Machine Intelligence,
More informationDecision Trees for Mining Data Streams Based on the Gaussian Approximation
International Journal of Computer Sciences and Engineering Open Access Review Paper Volume-4, Issue-3 E-ISSN: 2347-2693 Decision Trees for Mining Data Streams Based on the Gaussian Approximation S.Babu
More informationMore Local Structure Information for Make-Model Recognition
More Local Structure Information for Make-Model Recognition David Anthony Torres Dept. of Computer Science The University of California at San Diego La Jolla, CA 9093 Abstract An object classification
More informationNEW MEXICO Grade 6 MATHEMATICS STANDARDS
PROCESS STANDARDS To help New Mexico students achieve the Content Standards enumerated below, teachers are encouraged to base instruction on the following Process Standards: Problem Solving Build new mathematical
More informationClustering & Visualization
Chapter 5 Clustering & Visualization Clustering in high-dimensional databases is an important problem and there are a number of different clustering paradigms which are applicable to high-dimensional data.
More informationJiří Matas. Hough Transform
Hough Transform Jiří Matas Center for Machine Perception Department of Cybernetics, Faculty of Electrical Engineering Czech Technical University, Prague Many slides thanks to Kristen Grauman and Bastian
More informationA Genetic Algorithm-Evolved 3D Point Cloud Descriptor
A Genetic Algorithm-Evolved 3D Point Cloud Descriptor Dominik Wȩgrzyn and Luís A. Alexandre IT - Instituto de Telecomunicações Dept. of Computer Science, Univ. Beira Interior, 6200-001 Covilhã, Portugal
More informationParallel Data Selection Based on Neurodynamic Optimization in the Era of Big Data
Parallel Data Selection Based on Neurodynamic Optimization in the Era of Big Data Jun Wang Department of Mechanical and Automation Engineering The Chinese University of Hong Kong Shatin, New Territories,
More informationClassification of Fingerprints. Sarat C. Dass Department of Statistics & Probability
Classification of Fingerprints Sarat C. Dass Department of Statistics & Probability Fingerprint Classification Fingerprint classification is a coarse level partitioning of a fingerprint database into smaller
More informationBlog Post Extraction Using Title Finding
Blog Post Extraction Using Title Finding Linhai Song 1, 2, Xueqi Cheng 1, Yan Guo 1, Bo Wu 1, 2, Yu Wang 1, 2 1 Institute of Computing Technology, Chinese Academy of Sciences, Beijing 2 Graduate School
More informationINCIDENCE-BETWEENNESS GEOMETRY
INCIDENCE-BETWEENNESS GEOMETRY MATH 410, CSUSM. SPRING 2008. PROFESSOR AITKEN This document covers the geometry that can be developed with just the axioms related to incidence and betweenness. The full
More informationAutomatic Labeling of Lane Markings for Autonomous Vehicles
Automatic Labeling of Lane Markings for Autonomous Vehicles Jeffrey Kiske Stanford University 450 Serra Mall, Stanford, CA 94305 jkiske@stanford.edu 1. Introduction As autonomous vehicles become more popular,
More information2695 P a g e. IV Semester M.Tech (DCN) SJCIT Chickballapur Karnataka India
Integrity Preservation and Privacy Protection for Digital Medical Images M.Krishna Rani Dr.S.Bhargavi IV Semester M.Tech (DCN) SJCIT Chickballapur Karnataka India Abstract- In medical treatments, the integrity
More informationHow To Fix Out Of Focus And Blur Images With A Dynamic Template Matching Algorithm
IJSTE - International Journal of Science Technology & Engineering Volume 1 Issue 10 April 2015 ISSN (online): 2349-784X Image Estimation Algorithm for Out of Focus and Blur Images to Retrieve the Barcode
More informationCommon Core Unit Summary Grades 6 to 8
Common Core Unit Summary Grades 6 to 8 Grade 8: Unit 1: Congruence and Similarity- 8G1-8G5 rotations reflections and translations,( RRT=congruence) understand congruence of 2 d figures after RRT Dilations
More informationDYNAMIC DOMAIN CLASSIFICATION FOR FRACTAL IMAGE COMPRESSION
DYNAMIC DOMAIN CLASSIFICATION FOR FRACTAL IMAGE COMPRESSION K. Revathy 1 & M. Jayamohan 2 Department of Computer Science, University of Kerala, Thiruvananthapuram, Kerala, India 1 revathysrp@gmail.com
More informationA Genetic Algorithm Approach for Solving a Flexible Job Shop Scheduling Problem
A Genetic Algorithm Approach for Solving a Flexible Job Shop Scheduling Problem Sayedmohammadreza Vaghefinezhad 1, Kuan Yew Wong 2 1 Department of Manufacturing & Industrial Engineering, Faculty of Mechanical
More informationAutomatic 3D Reconstruction via Object Detection and 3D Transformable Model Matching CS 269 Class Project Report
Automatic 3D Reconstruction via Object Detection and 3D Transformable Model Matching CS 69 Class Project Report Junhua Mao and Lunbo Xu University of California, Los Angeles mjhustc@ucla.edu and lunbo
More informationComputational Geometry. Lecture 1: Introduction and Convex Hulls
Lecture 1: Introduction and convex hulls 1 Geometry: points, lines,... Plane (two-dimensional), R 2 Space (three-dimensional), R 3 Space (higher-dimensional), R d A point in the plane, 3-dimensional space,
More informationSubspace Analysis and Optimization for AAM Based Face Alignment
Subspace Analysis and Optimization for AAM Based Face Alignment Ming Zhao Chun Chen College of Computer Science Zhejiang University Hangzhou, 310027, P.R.China zhaoming1999@zju.edu.cn Stan Z. Li Microsoft
More informationEstimation of the COCOMO Model Parameters Using Genetic Algorithms for NASA Software Projects
Journal of Computer Science 2 (2): 118-123, 2006 ISSN 1549-3636 2006 Science Publications Estimation of the COCOMO Model Parameters Using Genetic Algorithms for NASA Software Projects Alaa F. Sheta Computers
More informationEmpirically Identifying the Best Genetic Algorithm for Covering Array Generation
Empirically Identifying the Best Genetic Algorithm for Covering Array Generation Liang Yalan 1, Changhai Nie 1, Jonathan M. Kauffman 2, Gregory M. Kapfhammer 2, Hareton Leung 3 1 Department of Computer
More informationA New Approach for Similar Images Using Game Theory
Applied Mathematical Sciences, Vol. 8, 2014, no. 163, 8099-8111 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.12988/ams.2014.410790 A New Approach for Similar Images Using Game Theory O. Bencharef
More informationChapter ML:XI (continued)
Chapter ML:XI (continued) XI. Cluster Analysis Data Mining Overview Cluster Analysis Basics Hierarchical Cluster Analysis Iterative Cluster Analysis Density-Based Cluster Analysis Cluster Evaluation Constrained
More informationPerformance of networks containing both MaxNet and SumNet links
Performance of networks containing both MaxNet and SumNet links Lachlan L. H. Andrew and Bartek P. Wydrowski Abstract Both MaxNet and SumNet are distributed congestion control architectures suitable for
More information3D Human Face Recognition Using Point Signature
3D Human Face Recognition Using Point Signature Chin-Seng Chua, Feng Han, Yeong-Khing Ho School of Electrical and Electronic Engineering Nanyang Technological University, Singapore 639798 ECSChua@ntu.edu.sg
More informationOptimization of PID parameters with an improved simplex PSO
Li et al. Journal of Inequalities and Applications (2015) 2015:325 DOI 10.1186/s13660-015-0785-2 R E S E A R C H Open Access Optimization of PID parameters with an improved simplex PSO Ji-min Li 1, Yeong-Cheng
More informationAlgebra 1 2008. Academic Content Standards Grade Eight and Grade Nine Ohio. Grade Eight. Number, Number Sense and Operations Standard
Academic Content Standards Grade Eight and Grade Nine Ohio Algebra 1 2008 Grade Eight STANDARDS Number, Number Sense and Operations Standard Number and Number Systems 1. Use scientific notation to express
More informationA Learning Based Method for Super-Resolution of Low Resolution Images
A Learning Based Method for Super-Resolution of Low Resolution Images Emre Ugur June 1, 2004 emre.ugur@ceng.metu.edu.tr Abstract The main objective of this project is the study of a learning based method
More informationARTICLE IN PRESS. European Journal of Operational Research xxx (2004) xxx xxx. Discrete Optimization. Nan Kong, Andrew J.
A factor 1 European Journal of Operational Research xxx (00) xxx xxx Discrete Optimization approximation algorithm for two-stage stochastic matching problems Nan Kong, Andrew J. Schaefer * Department of
More informationECE 533 Project Report Ashish Dhawan Aditi R. Ganesan
Handwritten Signature Verification ECE 533 Project Report by Ashish Dhawan Aditi R. Ganesan Contents 1. Abstract 3. 2. Introduction 4. 3. Approach 6. 4. Pre-processing 8. 5. Feature Extraction 9. 6. Verification
More informationThe Scientific Data Mining Process
Chapter 4 The Scientific Data Mining Process When I use a word, Humpty Dumpty said, in rather a scornful tone, it means just what I choose it to mean neither more nor less. Lewis Carroll [87, p. 214] In
More informationLinear Codes. Chapter 3. 3.1 Basics
Chapter 3 Linear Codes In order to define codes that we can encode and decode efficiently, we add more structure to the codespace. We shall be mainly interested in linear codes. A linear code of length
More informationA Method of Caption Detection in News Video
3rd International Conference on Multimedia Technology(ICMT 3) A Method of Caption Detection in News Video He HUANG, Ping SHI Abstract. News video is one of the most important media for people to get information.
More informationTHREE DIMENSIONAL GEOMETRY
Chapter 8 THREE DIMENSIONAL GEOMETRY 8.1 Introduction In this chapter we present a vector algebra approach to three dimensional geometry. The aim is to present standard properties of lines and planes,
More informationIN current film media, the increase in areal density has
IEEE TRANSACTIONS ON MAGNETICS, VOL. 44, NO. 1, JANUARY 2008 193 A New Read Channel Model for Patterned Media Storage Seyhan Karakulak, Paul H. Siegel, Fellow, IEEE, Jack K. Wolf, Life Fellow, IEEE, and
More informationSection 1.1. Introduction to R n
The Calculus of Functions of Several Variables Section. Introduction to R n Calculus is the study of functional relationships and how related quantities change with each other. In your first exposure to
More informationIntersection of a Line and a Convex. Hull of Points Cloud
Applied Mathematical Sciences, Vol. 7, 213, no. 13, 5139-5149 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/1.12988/ams.213.37372 Intersection of a Line and a Convex Hull of Points Cloud R. P. Koptelov
More informationCategorical Data Visualization and Clustering Using Subjective Factors
Categorical Data Visualization and Clustering Using Subjective Factors Chia-Hui Chang and Zhi-Kai Ding Department of Computer Science and Information Engineering, National Central University, Chung-Li,
More informationWii Remote Calibration Using the Sensor Bar
Wii Remote Calibration Using the Sensor Bar Alparslan Yildiz Abdullah Akay Yusuf Sinan Akgul GIT Vision Lab - http://vision.gyte.edu.tr Gebze Institute of Technology Kocaeli, Turkey {yildiz, akay, akgul}@bilmuh.gyte.edu.tr
More informationStochastic Processes and Queueing Theory used in Cloud Computer Performance Simulations
56 Stochastic Processes and Queueing Theory used in Cloud Computer Performance Simulations Stochastic Processes and Queueing Theory used in Cloud Computer Performance Simulations Florin-Cătălin ENACHE
More informationComputer Graphics. Geometric Modeling. Page 1. Copyright Gotsman, Elber, Barequet, Karni, Sheffer Computer Science - Technion. An Example.
An Example 2 3 4 Outline Objective: Develop methods and algorithms to mathematically model shape of real world objects Categories: Wire-Frame Representation Object is represented as as a set of points
More informationGA as a Data Optimization Tool for Predictive Analytics
GA as a Data Optimization Tool for Predictive Analytics Chandra.J 1, Dr.Nachamai.M 2,Dr.Anitha.S.Pillai 3 1Assistant Professor, Department of computer Science, Christ University, Bangalore,India, chandra.j@christunivesity.in
More informationEuler Vector: A Combinatorial Signature for Gray-Tone Images
Euler Vector: A Combinatorial Signature for Gray-Tone Images Arijit Bishnu, Bhargab B. Bhattacharya y, Malay K. Kundu, C. A. Murthy fbishnu t, bhargab, malay, murthyg@isical.ac.in Indian Statistical Institute,
More informationSegmentation of building models from dense 3D point-clouds
Segmentation of building models from dense 3D point-clouds Joachim Bauer, Konrad Karner, Konrad Schindler, Andreas Klaus, Christopher Zach VRVis Research Center for Virtual Reality and Visualization, Institute
More informationWe can display an object on a monitor screen in three different computer-model forms: Wireframe model Surface Model Solid model
CHAPTER 4 CURVES 4.1 Introduction In order to understand the significance of curves, we should look into the types of model representations that are used in geometric modeling. Curves play a very significant
More informationA New Method for Traffic Forecasting Based on the Data Mining Technology with Artificial Intelligent Algorithms
Research Journal of Applied Sciences, Engineering and Technology 5(12): 3417-3422, 213 ISSN: 24-7459; e-issn: 24-7467 Maxwell Scientific Organization, 213 Submitted: October 17, 212 Accepted: November
More informationLarge-Scale Data Sets Clustering Based on MapReduce and Hadoop
Journal of Computational Information Systems 7: 16 (2011) 5956-5963 Available at http://www.jofcis.com Large-Scale Data Sets Clustering Based on MapReduce and Hadoop Ping ZHOU, Jingsheng LEI, Wenjun YE
More informationSocial Media Mining. Data Mining Essentials
Introduction Data production rate has been increased dramatically (Big Data) and we are able store much more data than before E.g., purchase data, social media data, mobile phone data Businesses and customers
More informationModel-based Parameter Optimization of an Engine Control Unit using Genetic Algorithms
Symposium on Automotive/Avionics Avionics Systems Engineering (SAASE) 2009, UC San Diego Model-based Parameter Optimization of an Engine Control Unit using Genetic Algorithms Dipl.-Inform. Malte Lochau
More informationPre-Algebra 2008. Academic Content Standards Grade Eight Ohio. Number, Number Sense and Operations Standard. Number and Number Systems
Academic Content Standards Grade Eight Ohio Pre-Algebra 2008 STANDARDS Number, Number Sense and Operations Standard Number and Number Systems 1. Use scientific notation to express large numbers and small
More informationAUTOMATIC ADJUSTMENT FOR LASER SYSTEMS USING A STOCHASTIC BINARY SEARCH ALGORITHM TO COPE WITH NOISY SENSING DATA
INTERNATIONAL JOURNAL ON SMART SENSING AND INTELLIGENT SYSTEMS, VOL. 1, NO. 2, JUNE 2008 AUTOMATIC ADJUSTMENT FOR LASER SYSTEMS USING A STOCHASTIC BINARY SEARCH ALGORITHM TO COPE WITH NOISY SENSING DATA
More informationFast and Robust Normal Estimation for Point Clouds with Sharp Features
1/37 Fast and Robust Normal Estimation for Point Clouds with Sharp Features Alexandre Boulch & Renaud Marlet University Paris-Est, LIGM (UMR CNRS), Ecole des Ponts ParisTech Symposium on Geometry Processing
More informationFeature Selection using Integer and Binary coded Genetic Algorithm to improve the performance of SVM Classifier
Feature Selection using Integer and Binary coded Genetic Algorithm to improve the performance of SVM Classifier D.Nithya a, *, V.Suganya b,1, R.Saranya Irudaya Mary c,1 Abstract - This paper presents,
More informationSolving Geometric Problems with the Rotating Calipers *
Solving Geometric Problems with the Rotating Calipers * Godfried Toussaint School of Computer Science McGill University Montreal, Quebec, Canada ABSTRACT Shamos [1] recently showed that the diameter of
More informationDYNAMIC RANGE IMPROVEMENT THROUGH MULTIPLE EXPOSURES. Mark A. Robertson, Sean Borman, and Robert L. Stevenson
c 1999 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or
More informationDATA MINING CLUSTER ANALYSIS: BASIC CONCEPTS
DATA MINING CLUSTER ANALYSIS: BASIC CONCEPTS 1 AND ALGORITHMS Chiara Renso KDD-LAB ISTI- CNR, Pisa, Italy WHAT IS CLUSTER ANALYSIS? Finding groups of objects such that the objects in a group will be similar
More informationA Network Flow Approach in Cloud Computing
1 A Network Flow Approach in Cloud Computing Soheil Feizi, Amy Zhang, Muriel Médard RLE at MIT Abstract In this paper, by using network flow principles, we propose algorithms to address various challenges
More informationDetection and Restoration of Vertical Non-linear Scratches in Digitized Film Sequences
Detection and Restoration of Vertical Non-linear Scratches in Digitized Film Sequences Byoung-moon You 1, Kyung-tack Jung 2, Sang-kook Kim 2, and Doo-sung Hwang 3 1 L&Y Vision Technologies, Inc., Daejeon,
More informationHolland s GA Schema Theorem
Holland s GA Schema Theorem v Objective provide a formal model for the effectiveness of the GA search process. v In the following we will first approach the problem through the framework formalized by
More informationTaking Inverse Graphics Seriously
CSC2535: 2013 Advanced Machine Learning Taking Inverse Graphics Seriously Geoffrey Hinton Department of Computer Science University of Toronto The representation used by the neural nets that work best
More informationMulti-layer Structure of Data Center Based on Steiner Triple System
Journal of Computational Information Systems 9: 11 (2013) 4371 4378 Available at http://www.jofcis.com Multi-layer Structure of Data Center Based on Steiner Triple System Jianfei ZHANG 1, Zhiyi FANG 1,
More informationThe Essentials of CAGD
The Essentials of CAGD Chapter 2: Lines and Planes Gerald Farin & Dianne Hansford CRC Press, Taylor & Francis Group, An A K Peters Book www.farinhansford.com/books/essentials-cagd c 2000 Farin & Hansford
More informationSensors & Transducers 2015 by IFSA Publishing, S. L. http://www.sensorsportal.com
Sensors & Transducers 2015 by IFSA Publishing, S. L. http://www.sensorsportal.com A Dynamic Deployment Policy of Slave Controllers for Software Defined Network Yongqiang Yang and Gang Xu College of Computer
More informationStrictly Localized Construction of Planar Bounded-Degree Spanners of Unit Disk Graphs
Strictly Localized Construction of Planar Bounded-Degree Spanners of Unit Disk Graphs Iyad A. Kanj Ljubomir Perković Ge Xia Abstract We describe a new distributed and strictly-localized approach for constructing
More informationA Stock Pattern Recognition Algorithm Based on Neural Networks
A Stock Pattern Recognition Algorithm Based on Neural Networks Xinyu Guo guoxinyu@icst.pku.edu.cn Xun Liang liangxun@icst.pku.edu.cn Xiang Li lixiang@icst.pku.edu.cn Abstract pattern respectively. Recent
More informationA Method of Cloud Resource Load Balancing Scheduling Based on Improved Adaptive Genetic Algorithm
Journal of Information & Computational Science 9: 16 (2012) 4801 4809 Available at http://www.joics.com A Method of Cloud Resource Load Balancing Scheduling Based on Improved Adaptive Genetic Algorithm
More informationVision based Vehicle Tracking using a high angle camera
Vision based Vehicle Tracking using a high angle camera Raúl Ignacio Ramos García Dule Shu gramos@clemson.edu dshu@clemson.edu Abstract A vehicle tracking and grouping algorithm is presented in this work
More informationMachine Learning for Medical Image Analysis. A. Criminisi & the InnerEye team @ MSRC
Machine Learning for Medical Image Analysis A. Criminisi & the InnerEye team @ MSRC Medical image analysis the goal Automatic, semantic analysis and quantification of what observed in medical scans Brain
More informationParallel Computing for Option Pricing Based on the Backward Stochastic Differential Equation
Parallel Computing for Option Pricing Based on the Backward Stochastic Differential Equation Ying Peng, Bin Gong, Hui Liu, and Yanxin Zhang School of Computer Science and Technology, Shandong University,
More informationEvolutionary Detection of Rules for Text Categorization. Application to Spam Filtering
Advances in Intelligent Systems and Technologies Proceedings ECIT2004 - Third European Conference on Intelligent Systems and Technologies Iasi, Romania, July 21-23, 2004 Evolutionary Detection of Rules
More informationTwo novel characteristics in palmprint verification: datum point invariance and line feature matching
Pattern Recognition 32 (1999) 691 702 Two novel characteristics in palmprint verification: datum point invariance and line feature matching Dapeng Zhang*, Wei Shu Department of Computing, Hong Kong Polytechnic
More informationChapter 6. The stacking ensemble approach
82 This chapter proposes the stacking ensemble approach for combining different data mining classifiers to get better performance. Other combination techniques like voting, bagging etc are also described
More informationMedical Information Management & Mining. You Chen Jan,15, 2013 You.chen@vanderbilt.edu
Medical Information Management & Mining You Chen Jan,15, 2013 You.chen@vanderbilt.edu 1 Trees Building Materials Trees cannot be used to build a house directly. How can we transform trees to building materials?
More informationVector storage and access; algorithms in GIS. This is lecture 6
Vector storage and access; algorithms in GIS This is lecture 6 Vector data storage and access Vectors are built from points, line and areas. (x,y) Surface: (x,y,z) Vector data access Access to vector
More informationAPPLYING COMPUTER VISION TECHNIQUES TO TOPOGRAPHIC OBJECTS
APPLYING COMPUTER VISION TECHNIQUES TO TOPOGRAPHIC OBJECTS Laura Keyes, Adam Winstanley Department of Computer Science National University of Ireland Maynooth Co. Kildare, Ireland lkeyes@cs.may.ie, Adam.Winstanley@may.ie
More informationJunghyun Ahn Changho Sung Tag Gon Kim. Korea Advanced Institute of Science and Technology (KAIST) 373-1 Kuseong-dong, Yuseong-gu Daejoen, Korea
Proceedings of the 211 Winter Simulation Conference S. Jain, R. R. Creasey, J. Himmelspach, K. P. White, and M. Fu, eds. A BINARY PARTITION-BASED MATCHING ALGORITHM FOR DATA DISTRIBUTION MANAGEMENT Junghyun
More informationMultiobjective Multicast Routing Algorithm
Multiobjective Multicast Routing Algorithm Jorge Crichigno, Benjamín Barán P. O. Box 9 - National University of Asunción Asunción Paraguay. Tel/Fax: (+9-) 89 {jcrichigno, bbaran}@cnc.una.py http://www.una.py
More informationUsing Lexical Similarity in Handwritten Word Recognition
Using Lexical Similarity in Handwritten Word Recognition Jaehwa Park and Venu Govindaraju Center of Excellence for Document Analysis and Recognition (CEDAR) Department of Computer Science and Engineering
More informationA Parallel Processor for Distributed Genetic Algorithm with Redundant Binary Number
A Parallel Processor for Distributed Genetic Algorithm with Redundant Binary Number 1 Tomohiro KAMIMURA, 2 Akinori KANASUGI 1 Department of Electronics, Tokyo Denki University, 07ee055@ms.dendai.ac.jp
More informationThe Trip Scheduling Problem
The Trip Scheduling Problem Claudia Archetti Department of Quantitative Methods, University of Brescia Contrada Santa Chiara 50, 25122 Brescia, Italy Martin Savelsbergh School of Industrial and Systems
More informationARTIFICIAL INTELLIGENCE (CSCU9YE) LECTURE 6: MACHINE LEARNING 2: UNSUPERVISED LEARNING (CLUSTERING)
ARTIFICIAL INTELLIGENCE (CSCU9YE) LECTURE 6: MACHINE LEARNING 2: UNSUPERVISED LEARNING (CLUSTERING) Gabriela Ochoa http://www.cs.stir.ac.uk/~goc/ OUTLINE Preliminaries Classification and Clustering Applications
More informationOPTIMIZED SENSOR NODES BY FAULT NODE RECOVERY ALGORITHM
OPTIMIZED SENSOR NODES BY FAULT NODE RECOVERY ALGORITHM S. Sofia 1, M.Varghese 2 1 Student, Department of CSE, IJCET 2 Professor, Department of CSE, IJCET Abstract This paper proposes fault node recovery
More informationA new cut detection algorithm with constant false-alarm ratio for video segmentation
J. Vis. Commun. Image R. 5 (004) 44 www.elsevier.com/locate/jvci A new cut detection algorithm with constant false-alarm ratio for video segmentation Tie-Yan Liu, a Kwok-Tung Lo, b, * Xu-Dong Zhang, a
More information. Learn the number of classes and the structure of each class using similarity between unlabeled training patterns
Outline Part 1: of data clustering Non-Supervised Learning and Clustering : Problem formulation cluster analysis : Taxonomies of Clustering Techniques : Data types and Proximity Measures : Difficulties
More informationAdaptive Linear Programming Decoding
Adaptive Linear Programming Decoding Mohammad H. Taghavi and Paul H. Siegel ECE Department, University of California, San Diego Email: (mtaghavi, psiegel)@ucsd.edu ISIT 2006, Seattle, USA, July 9 14, 2006
More informationDecision-making with the AHP: Why is the principal eigenvector necessary
European Journal of Operational Research 145 (2003) 85 91 Decision Aiding Decision-making with the AHP: Why is the principal eigenvector necessary Thomas L. Saaty * University of Pittsburgh, Pittsburgh,
More informationFor example, estimate the population of the United States as 3 times 10⁸ and the
CCSS: Mathematics The Number System CCSS: Grade 8 8.NS.A. Know that there are numbers that are not rational, and approximate them by rational numbers. 8.NS.A.1. Understand informally that every number
More informationCurrent Standard: Mathematical Concepts and Applications Shape, Space, and Measurement- Primary
Shape, Space, and Measurement- Primary A student shall apply concepts of shape, space, and measurement to solve problems involving two- and three-dimensional shapes by demonstrating an understanding of:
More informationhttp://www.elsevier.com/copyright
This article was published in an Elsevier journal. The attached copy is furnished to the author for non-commercial research and education use, including for instruction at the author s institution, sharing
More information